1. Clinical Nutrition Experimental
Model-based analysis of postprandial glycemic response dynamics for different types of food 
Yvonne J. Rozendaal, Anne H. Maas, Carola van Pul, Eduardus J. Cottaar, Harm R. Haak, Peter A. Hilbers, Natal A. van Riel 
Clinical Nutrition Experimental 
Volume 19, Pages (June 2018)
DOI: /j.yclnex
Copyright © 2018 The Authors Terms and Conditions

2. Fig. 1
Schematic visualization of the mathematical model describing postprandial glucose metabolism. The mathematical model comprises three compartments (shaded background) depicted in the central column that represent the compartments in which the glucose and insulin balances are computed. The corresponding glucose and insulin fluxes (exchange between compartments) are visualized in dark gray and light gray, respectively, and are labeled with the model parameters that govern these fluxes adapted with permission from [28].
Clinical Nutrition Experimental  , 32-45DOI: ( /j.yclnex )
Copyright © 2018 The Authors Terms and Conditions

3. Fig. 2
Distribution of postprandial response data for all 53 included food products and meals. In this density plot, data for glucose (a) and insulin (b) are visualized over time for each included food product and meal [6–8,32–46], originating from in total 240 subjects. The data points (dots) are color-coded such that a richer shade corresponds to more datasets overlapping in this time–concentration region. The solid lines indicate the time course of the postprandial response profile for each food product and meal separately.
Clinical Nutrition Experimental  , 32-45DOI: ( /j.yclnex )
Copyright © 2018 The Authors Terms and Conditions

4. Fig. 3
How to quantify postprandial response dynamics? Panel a–c present currently available measures of postprandial response dynamics. In panel a datasets (black; error bars represent mean ± standard deviation per food and per time point) comprising of 50 g carbohydrate containing food are depicted, ranging from plain bread, fruit juice and rice up to complete breakfasts [7,8,33,35,39,43]. The gray area depicts the large range in which these postprandial responses lie. Panel b depicts datasets with similar 2 h-iAUC values assessed from the average glucose data for a cheese omelet with bread meal (black) [44] and a low GI snack (gray) [40]. The 2 h-iAUC is approximated using trapezoidal numerical integration (conform current standards [62], only the incremental area – above fasting level – is included). Panel c presents the assessed kinetic properties for two different types of food both having a Glycemic Index of 70 and containing 50 g of digestible carbohydrate: white bread [7] is shown in black, cornstarch [43] in gray. Panel d depicts the dependency on sampling frequency. The kinetic properties are assessed for the postprandial glucose profile for a high GI breakfast containing 65 g digestible carbohydrates. The original dataset [38] is shown in black, and in gray a subset of the same dataset is shown that has a reduced temporal resolution and shorter measurement duration. in panel e, spline fitting (gray) is examined in the case of a wheat lunch [36] (data shown in black). Panel f illustrates the physiology-based dynamic modeling approach in terms of how the underlying fluxes regulate glucose and insulin in the plasma during the postprandial state. Panel g illustrates the kinetic properties to quantify the characteristic dynamics of postprandial profiles.
Clinical Nutrition Experimental  , 32-45DOI: ( /j.yclnex )
Copyright © 2018 The Authors Terms and Conditions

5. Fig. 4
The physiology-based dynamic model describes the heterogeneity in postprandial dynamics well. Panels a–b show the postprandial response for various equi-carbohydrate foods and panels c–d for foods with varying carbohydrate content. Simulated (solid lines) and observed (error bars: mean ± standard deviation) postprandial glucose (a,c) and insulin (b,d) response profiles for a selection of food products and meals. The OGTT model simulation (a–b) is included in black to serve as reference. Panels a-b comprise situations in which 50 g available carbohydrates are present in the ingested food [7,8,39,43], but result in different postprandial dynamics. Panels c–d depict the opposite situation in which the postprandial dynamics falls in a similar range, although the available carbohydrate content varies [40,45].
Clinical Nutrition Experimental  , 32-45DOI: ( /j.yclnex )
Copyright © 2018 The Authors Terms and Conditions

6. Fig. 5
The physiology-based dynamic model predicts the underlying metabolic fluxes. The predicted fluxes are displayed for foods that all contain 50 g of available carbohydrates [7,8,39,43]: rate of appearance of exogenous glucose (a), insulin synthesis rate (b) and glucose utilization by insulin-dependent tissues (c). These modeled fluxes correspond to the simulated glucose and insulin profiles depicted in Fig. 4a–b.
Clinical Nutrition Experimental  , 32-45DOI: ( /j.yclnex )
Copyright © 2018 The Authors Terms and Conditions

7. Fig. 6
The physiology-based dynamic model can also describe postprandial profiles for pre-diabetic cases. Panel a shows the postprandial glucose level whereas panel b describes the postprandial insulin level. The black error bars represent the data by Nazare et al. [38] following a low GI meal in non-diabetic overweight subjects. The solid lines represent the model simulations with the healthy model parameters for beta-cell function and insulin sensitivity. The dashed lines show the model simulations using the adapted model parameters where the insulin sensitivity and beta cell function are re-estimated to describe the pre-diabetic case more closely.
Clinical Nutrition Experimental  , 32-45DOI: ( /j.yclnex )
Copyright © 2018 The Authors Terms and Conditions